TARGET_ASP_RATIO = 16.0/9

haar_face_cascade = cv2.CascadeClassifier('./haarcascade_frontalface_default.xml')

# haar_face_cascade = cv2.CascadeClassifier('./haarcascade_glasses.xml')

haar_eye_cascade = cv2.CascadeClassifier('./haarcascade_eye.xml')

def crop_to_first_face(image_copy, faces_rect, debug=False):

    # Logic to detect cropping area

    y_crop = 0;

    crop_height = int(image_copy.shape[1]*1.0/TARGET_ASP_RATIO)

    if debug:

        print("crop_height: " + str(crop_height))

    # Pick the first face from top and create boundary

    if (len(faces_rect) > 0):

        sorted_faces_rect = sorted(faces_rect , key = lambda x: float(x[1]))

        x, y, w, h = sorted_faces_rect[0]

        if debug:

            print("Face Dims: ")

            print(x,y,w,h)

        # Start crop from double the face height

        yc = int(y - int(0.5*h))

        if yc < 0:

            yc = 0;

        y_crop = yc

        if debug:

            print("Crop y positions: ")

            print(y_crop, y_crop+crop_height)

        # Create a border to highlight image cropping

        cv2.rectangle(image_copy, (0, y_crop), 

                      (image_copy.shape[1], y_crop+crop_height), 

                      (0, 0, 250), 30)

        # Crop the image

        crop_img = image_copy[y_crop:y_crop+crop_height, 0:image_copy.shape[1]]

        if debug:

            print("crop_img: ")

            print(crop_img.shape)

    else:

        # no faces found, rotate the image

        crop_img = cv2.rotate(image_copy, cv2.ROTATE_90_CLOCKWISE)

        cv2.rectangle(crop_img, (0,0), (crop_img.shape[1], crop_img.shape[0]),

                      (0, 250, 0), 30)

​

    return crop_img

def detect_faces(cascade, test_image, debug=False):

    # create a copy of the image to prevent any changes to the original one.

    image_copy = test_image.copy()

​

    #convert the test image to gray scale as opencv face detector expects gray images

    gray_image = cv2.cvtColor(image_copy, cv2.COLOR_BGR2GRAY)

​

    # Applying the haar classifier to detect faces

    faces_rect = cascade.detectMultiScale(gray_image, scaleFactor=1.1, 

                                          minNeighbors=5, minSize=(100,100))

​

    if debug:

        print("faces_rect: " + str(len(faces_rect)))

​

#     for (x, y, w, h) in faces_rect:

#         cv2.rectangle(image_copy, (x, y), (x+w, y+h), (255, 0, 0), 2)    

    crop_img = crop_to_first_face(image_copy, faces_rect, debug)

#     crop_to_most_faces(image_copy, faces_rect)

    return crop_img

​

def convertToRGB(image):

    return cv2.cvtColor(image, cv2.COLOR_BGR2RGB)

# loading image

# test_image = cv2.imread('/Users/amitsharma/Downloads/RajasthanTrip-Dec2019/20191230_160138.jpg')

test_image = cv2.imread('/Users/amitsharma/axs/hobby/raspi/slideshow/output/Portrait/20200126_092939.jpg')

​

​

# test_image = cv2.imread('/Users/amitsharma/Downloads/IMG-20160214-WA002.jpg')

​

​

#call the function to detect faces

faces = detect_faces(haar_face_cascade, test_image, debug=True)

​

fig=plt.figure()

# fig.add_subplot(1,1,1)

plt.imshow(convertToRGB(test_image))

​

#convert to RGB and display image

# plt.figure(figsize = (10,10))

# fig.add_subplot(1,2,2)

fig=plt.figure(figsize=(12,12))

plt.imshow(convertToRGB(faces))

PORTRAIT_DIR = "/Users/amitsharma/axs/hobby/raspi/slideshow/output/Portrait/"

os.chdir(PORTRAIT_DIR)

for file in glob.glob("*.jpg"):

    print("Processing: " + PORTRAIT_DIR + file)

    # loading image

    test_image = cv2.imread(PORTRAIT_DIR + file)

    fig=plt.figure(1)

    # Original image in left side

    fig.add_subplot(1,2,1)

    plt.imshow(convertToRGB(test_image))

​

    # crop the image

    crop_img = detect_faces(haar_face_cascade, test_image)

    # Cropped image on right side

    fig.add_subplot(1,2,2)

    plt.imshow(convertToRGB(crop_img))

    plt.show()